OpenAI collaborates with Broadcom to produce an AI chip for internal operations

OpenAI is reportedly preparing to deploy its first in-house AI chip in collaboration with Broadcom in 2026. This new chip is designed exclusively for internal applications and will optimise ChatGPT-scale model operations.

The move comes as OpenAI seeks to lessen dependence on Nvidia, a dominant supplier in the AI chip market. By controlling the hardware, OpenAI can achieve better efficiency, reduce operational costs, and improve computing performance for large-scale models.

The Road Ahead for OpenAI’s AI Infrastructure

Reports suggest that the chip design has been finalised in partnership with Broadcom and TSMC, which will manufacture the silicon. OpenAI’s focus on internal deployment ensures that the hardware will be used solely to accelerate its AI services rather than being sold to external clients.

Broadcom CEO Hock Tan disclosed that the company has received over $10 billion in AI infrastructure orders from a new customer, widely speculated to be OpenAI. Custom silicon solutions are expected to cater to the specific workload demands of large language models, highlighting the importance of hardware optimisation in AI research.

Custom chips are increasingly common among technology leaders. Companies like Google, Meta, and Amazon have already invested in building their own AI hardware to meet growing computational needs. OpenAI’s initiative positions it alongside these companies while providing control over performance, energy consumption, and design flexibility.

Internal use enables OpenAI to test new AI architectures, enhance processing efficiency, and explore innovative AI applications. The company can optimise speed, memory utilisation, and processing capacity for its large models, providing a competitive edge in AI development.

The partnership with Broadcom exemplifies a growing trend in the AI sector, where companies are aiming to integrate hardware and software more closely to enhance reliability and efficiency. Industry experts suggest that such moves could reshape AI research infrastructure and supply chains. OpenAI and Broadcom have not shared technical specifications or launch timelines publicly. The collaboration reflects the strategic importance of having tailored hardware for next-generation AI models and shows OpenAI’s commitment to strengthening its computational capabilities ahead of future innovations.